With recent advancements in augmented reality (AR) display technologies, various head-mounted AR displays are now on the consumer market (e.g., Microsoft HoloLens 2, Meta Quest 3). However, there is limited knowledge on how these modern displays impact depth perception. Moreover, understanding the significance of occlusion's effects on action-space distance perception in AR is important, particularly for practical applications such as AR-assisted search-and-rescue operations. However, to our knowledge, no prior research has examined how varying occluder surfaces affect action-space distance estimation across head-mounted AR devices (optical see-through (OST) and video see-through (VST)). In this paper, we present the first research to examine how different occluding surfaces affect action-space distance estimation using OST and VST head-mounted AR devices. With 28 participants across two experiments, we examined verbal estimates and blind walking distance estimation methods. Three occluding surfaces: a checkerboard (high-salient), a black surface (low-salient), and a realistic brick wall texture were considered. Five action-space distances, ranging from 1m to 5m, were evaluated using two commercial AR devices: Microsoft HoloLens 2 (OST) and Meta Quest 3 (VST). Our findings suggest that occluders do not impact AR object action-space distance estimations through blind walking and verbal distance estimation tasks with two modern AR devices. Blind walking distance estimations varied between the HoloLens 2 and Meta Quest 3, with the HoloLens 2 (OST) achieving higher accuracy (distance error: 0.4cm to 3m), than the Meta Quest 3 (VST) (distance error: 24cm to 3m), regardless of occlusion. For AR objects placed on or behind an occluder, distance estimations were overestimated with Meta Quest 3, while the HoloLens 2 either achieved close to accurate or underestimated of depth perception via blind walking.

A primary objective of optical see-through (OST) Augmented Reality (AR) systems is to enable users to perceive the depth of AR objects with the same accuracy as they perceive the depth of realworld objects. Previous studies have shown that individuals often make depth judgments that either underestimate or overestimate the depth of AR objects compared to real-world objects. Recently, OST AR devices have incorporated eye-tracking technology, offering the opportunity to objectively measure and investigate the depth-dependent components of the human visual system (e.g., eye vergence angle) while perceiving the depth of real and AR objects. This paper measures and examines the eye vergence angle (EVA) for both real and AR objects at four different depths (0.35 m, 0.75 m, 1.5 m, and 4.0 m) using integrated eye-tracking systems of the two commercial devices: Microsoft HoloLens 2 and Magic Leap 2. The experiment considered a four-alternative forced-choice visual discrimination task with a repeated-measures design, involving 24 participants. Our findings showed that subjective (verbal estimation) and objective (EVA) measures of depth perception were consistent with the depth of both real and AR objects in both OST AR devices. The results demonstrated individual differences in EVA for each device across various depths. No difference in EVA was identified between real and AR objects. Additionally, the EVA range differs between the OST AR devices for each depth. These findings indicate that objective EVA measurements can not be generalized across different OST AR devices and individuals.